Partial least squares modeling of retention data of oxo compounds in gas chromatography

Partial least squares modeling of latent structures were carried out on a d ata matrix consisting of Kovats retention indices of 35 aliphatic ketones a nd aldehydes and their physical characteristics. The retention indices were determined on capillary columns with 4 different stationary phases, namely bonded methyl- {HP-1}, methyl-phenyl- {HP-50}, and trifluoropropyl-methyl siloxane {DB-210}, as well as polyethylene glycol {HP-Innowax} at four diff erent temperatures. It was found that ketones and aldehydes cannot be class ified on the basis of retention data solely whereas the physical characteri stics (boiling point, molar volume, molecular mass, molar refraction, octan ol-water partition coefficient) contain the necessary information for diffe rentiation of the two classes of compounds. The retention index of but-2-en al does not fit into the general trend of aliphatic aldehydes and ketones. A predictive PLS model was built to estimate retention data of oxo compound s at different temperatures and various stationary phases of different pola rity. Cross-validation suggests a good reliability of the results. Characte ristic plots (PLS weights, scores) show the similar retention behavior of o xo compounds (Figures 3 and 4).